Adjuvant systemic treatment for resectable breast cancer changes the natural history of the disease but provides only a small and delayed effect on survival. Evaluation of the types of first relapse avoided by available treatments may explain why effects on mortality are small and appear late during follow-up. In randomised clinical trials done by the International Breast Cancer Study Group (IBCSG) between 1978 and 1985, 2108 patients with node-positive disease received more-effective treatments (6 or more cycles of cyclophosphamide, methotrexate, fluorouracil and prednisone; with or without tamoxifen, or tamoxifen and prednisone alone), and 722 patients received less-effective treatments (no treatment or a single cycle of chemotherapy). 3 main categories of first site of relapse were defined and evaluated by cumulative incidence analysis: local or regional, and distant soft tissue, bone, and viscera. The more-effective treatments reduced the cumulative incidence of first relapse in local or regional and distant soft tissue sites at 10 years from 36% to 18% (p = 0.0001); first relapse in bone and viscera was not altered by the more-effective treatments. These results were similar for premenopausal and postmenopausal women, and for patients with oestrogen-receptor-positive or oestrogen-receptor-negative tumours. Adjuvant systemic treatments in current use improve patient outcome mainly by reducing the incidence of first local or regional and distant soft-tissue relapses, while first recurrences in bone or viscera are influenced much less. More intensive treatments at present being tested in clinical trials might affect bone and visceral relapses and have a greater and earlier influence on survival.